>> From Flex, my questions in red:
>> "A number of people have asked how you can get more than 96 dB of
> instantaneous dynamic range out of a 16-bit A/D converter. You may think
> that one can only achieve 6 dB per bit, which would be 96 dB. Technically
> the theoretical maximum limit is 6.02n +1.67 dB (wheren is the number of
> bits).[1,2] What many people fail to understand is that dynamic range is a
> meaningless term without knowing the final detection bandwidth (i.e. 500 Hz
> CW filter). Indeed narrowing the final detection bandwidth reduces noise as
> well. That was true with analog filters too. The only problem with this
> approach in analog signal filtering was that added electronic again was
> required to offset the filter loss, often with an INCREASE in overall noise
> due to the added noise from the amplifier’s noise figure. DSP filtering
> does not suffer in this regard.
>>
>> Instantaneous dynamic range increases with decreasing bandwidth by a
> factor of 10*log*(bandwidth change). That means that a 50 Hz filter will
> provide 10 dB higher dynamic range than a 500 Hz filter. That is why you
> hear less noise in the smaller filter. The actual receiver noise figure
> (NF) of the radio has not changed but the detection bandwidth has. Thus
> the SNR and dynamic range improves accordingly. Within the bandwidth being
> detected! Therefore, yes, the dynamic range does improve at narrower
> bandwidths if the dynamic range is noise limited. This would, of course, be
> true on 40 meters in the summer. However, unless I messed something, this
> answer does not address the problems that take place when a signal whose
> amplitude is 148 dB above the LSB (least significant bit) voltage. An A/D
> converter has a voltage change for each bit. If the span of the A/D is 10
> volts, then each bit in a 16 bit A/D accounts for a change in amplitude of
> 10V/65536 = 0.0015 V/bit (1.5 mV/bit). If you assume that a signal on the
> antenna of 1 uV is “copyable” in CW, and is thus at least a few dB above
> the noise floor, S1, for example, on a quiet band, then there would need to
> be some sort of gain stage (~1000X or 30 dB) to bring this up to one least
> significant bit. Alternatively, you design so that the LSB step size is
> 1/1000 of 10 volts or 10 mV. That is allowing only 1 bit to give you your
> CW signal. Realistically you need 4 bits to give a readable signal on CW,
> about 6 to 8 bits for SSB. So when you are detecting this 1 uV signal with
> a 50 Hz filter in place, all is well and you have good instantaneous
> dynamic range.
>>
>>
>>
>> However, the A/D is simultaneously looking at the entire HF spectrum all
> of the time. It is digitizing that entire spectrum all of the time at twice
> (or more) the Nyquist frequency, and is extracting each individual signal
> using the DSP to unravel the complex waveform (which is the Fourier
> transform of the entire waveform that is digitized). If a signal appears
> anywhere between 0 and 30 MHz with an amplitude greater than 96 dB above 1
> uV, with a 16 bit A/D, you have a problem. A signal 148 dB greater than 1
> uV, quite often the case for the large shortwave station down the block,
> presents a signal voltage of 25 volts. Houston, we have a problem.
>>
>> The law of linear superposition, essential for any of this stuff to
> work, simply states that the instantaneous value of any sampled point is
> the linear sum of all of the components at the sampling point. Thus, the 25
> volt signal from that shortwave station rides on top of the 1 uV signal you
> are trying to hear. In reality, this will be well outside the“ maximum peak
> signal handling capability” of the A/D converter.
>>
>>
>>
>> When you get outside the“ maximum peak signal handling capability” of
> the A/D converter, the law of linear superposition is still true, but the
> A/D cannot measure that signal because it is out of range. Very non-linear
> things happen when an A/D converter gets outside the“ maximum peak signal
> handling capability”. That would be known in analog days as “overload” or
> “blocking”.
>>
>>
>>
>> The dynamic range of any ADC is normally assumed to be specified over
> the Nyquist bandwidth, which is equal to 1/2 of the converter’s sampling
> rate. With the ADC used in the FLEX-6000 series, the Nyquist bandwidth is
> 122.88 MHz. To calculate instantaneous dynamic range, one needs to know
> the converter’s specified signal to noise ratio (SNR), maximum peak signal
> handling capability, sampling rate, and final detection bandwidth. There
> are many application notes available from Analog Devices, Linear
> Technology, Texas Instruments, etc. that aid in these calculations. It is
> beyond the scope of this newsletter to provide the detailed education and
> analysis."
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|